[gmx-users] trjconv, g_mindist and others in parallel

[Erik Lindahl]

Hi Mike,

A couple of the computationally heavy tools are parallel (concoord,  
etc), but most of them are not.

The problem is that analysis is usually limited by disk I/O throughput,  
so adding more CPUs won't help.

One way to solve it is to split your trajectory in multiple shorter  
ones, and then run the analysis in parallel on nodes with different  
disk systems.

Cheers,

Erik

On Thursday, Apr 24, 2003, at 10:31 US/Pacific, Mike Tallhammer wrote:

> I was wondering if one could run trjconv, g_mindist and other GROMACS
> utilities in parallel if you have the parallel version compiled on a  
> dual
> processor machine. I'm trying to run some analysis on a trajectory of
> approx. 70 GB and running in parallel would be of great gain to me.  
> Any help
> would be greatly appreciated.
>
> Thanks
> Michael Tallhamer
> Asst. Researcher
> Cleveland State University
> Cleveland, OH 44115
> E-Mail: m.tallhamer at csuohio.edu
>
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
>
------------------------------------------------------------------------ 
-----
Erik Lindahl, MSc, PhD     <lindahl at stanford.edu>
D109, Fairchild Building
Dept. Structural Biology, Stanford University School of Medicine
Tel. 650-7250754    Fax. 650-7238464